Today's Workshop

• Introduction
• Data Fetching
  • requests
• Data Manipulating
  • numpy, pandas
• Data Visualizations
  • matplotlib

Introduction

We will explore tools in Python that will allow us to explore hockey data

Goal: Visualize where on the field skaters took shots and scored

... but to do anything, we need a data source

Data Sources

Big Data Cup

Kaggle

Specialty sources

Their Hockey Counts

CHL microstats

NHL microstats

But today we will be gathering data straight from the source - NHL API

NHL API

http://statsapi.web.nhl.com/api/v1/game/2021020002/feed/live

Three Parts the the Game ID*

• Year: 2021
• Season: 02
• Game Number: 0002

NHL API

import requests
year = '2021'
season_type = '02' 
max_game_ID = 2000

NHL API

def get_game(i):
    game_ID = year + season_type +str(i).zfill(4)
    try:
        r = requests.get(url='http://statsapi.web.nhl.com/api/v1/game/'+ game_ID +'/feed/live',
                         timeout=5)
        data = r.json()
        return data
    except:
        pass

Speeding it up with multiprocessing

import multiprocessing
with multiprocessing.Pool(20) as pool:
        game_Data = pool.map(get_game, range(1,max_game_ID))

Extracting Data from the request

# event, x, y, team_o (offence team), team_d (defending team), period, player_o, year, month, time_in_period, date_time
master_data = []
for data in game_data:
    teams = {
        "away": data['gameData']['teams']['away']['name'],
        "home": data['gameData']['teams']['home']['name']
    }
    teams_list = list(teams.values())
    plays = data['liveData']['plays']['allPlays']
    # Year and month of game
    date = data['gameData']['datetime']['dateTime'].split('-')
    year = date[0]
    month = date[1]

Extracting Data from the request

We care of two event types- Shot and Goal

    for play in plays:
        for event in event_types:
            if play['result']['event'] in [event]:
                if 'x' in play['coordinates'] and 'y' in play['coordinates']:
                    x = play['coordinates']['x']
                    y = play['coordinates']['y']
                TEAM = play['team']['name']
                if TEAM:
                    team_o = TEAM
                    team_d = teams_list[(teams_list.index(TEAM) + 1) % 2]
                # period
                period = play['about']['period']
                master_data.append([event, x, y, team_o, team_d, period, player_o, year, month])

Data Manipulation

Once you have all the data, you can convert it into a pandas DataFrame for easier analysis

df = pd.DataFrame(master_data, columns=['event', 'x', 'y', 'team_o', 'team_d', 'period', 'player_o', 'year', 'month', 'time', 'datetime'])

Now the Fun part!

Now that we have all the data in the form we want, we can now start playing with the data!

Lets say I want to rank skaters with the highest scoring average ... What should I do?

Visualizing Goals

Lets say I want to see where all the Goals scored on Tampa Bay Lightning were shot from, I can do a quick pandas filter and get all the data I need

coors = df[ (df['team_d'] == 'Tampa Bay Lightning') & (df['event'] == 'Goal') ][['x', 'y']]

Improved Visualization - Code

Place a picture of the field underneath the scatter plot

from PIL import Image, ImageOps
image_file = "Hockey-field-half.png"
image = Image.open(image_file).resize((400,340))
image_arr = np.array(image.getdata())
image_arr = image_arr.reshape(image.size[1], image.size[0], 4)